Much of the time involved in model parsing can be avoided if a similar model is supplied. Might even be possible to automatically cache models